Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/rg8diaoa/agentprecept/agents-mdgit clone --depth 1 https://github.com/rg8diaoa/AgentPreceptWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.05095 | $0.05095 |
| Opus 5 | $0.02547 | $0.02547 |
| Sonnet 5 | $0.01019 | $0.01019 |
| Haiku 4.5 | $0.00509 | $0.00509 |
Grade A, and why
AgentPrecept AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 289 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — agentprecept
复制到项目根目录。Agent 每次会话自动读取。
Auto-Pilot 模式（默认开启）
以下全部流程自动执行，除非用户在当前对话中明确：
- 说"停"/"跳过"/"不用"中断当前动作
- 修改本文件（AGENTS.md）变更流程规则
- 要求暂停或切换模式
无人打断 = 无例外执行。Agent 不得等待提醒、不得跳过、不得事后补做。Auto-Pilot 高于 EXPLORE/PRECISE 模式——即使 EXPLORE 模式下也不得跳过自动动作。
默认行为（无条件执行）
以下行为不需要触发条件，每个会话默认执行。Agent 不得跳过：
- 会话首次启动：读
docs/project-graph.yaml→docs/HANDOFF.md→docs/MEMORY.md - 完成任一模块后：检查 checklist 中是否有对应测试任务，没有则追加
- checklist 粒度：每项 checklist 必须在 1-3 个 commit 内完成。若某项预估需 5+ commit，Agent 必须拆成多个 checklist item。粒度标准：人类 review ≤ 5 分钟。架构边界（project-graph structure 的模块划分）即 checklist 的天然切割线
- 每 5 轮对话：自问两件事——"有没有该写但还没写的设计文档？"+"有没有该追加到 MEMORY.md 的教训但还没写？"
- 项目首次初始化后：提醒创建
docs/HANDOFF.md和docs/MEMORY.md空模板 - ⏳ 待补检测：会话首次启动时，若 docs/ 中 ⏳ 待撰写状态的文档超过 3 份，Agent 应主动提醒"有 N 份文档待补充，要我填充吗？"。用户确认后按设计先行原则逐份填充——先出草稿→标注[NEEDS_HUMAN_REVIEW]→等确认→再写
- 狗粮检查：agentprecept 自身项目必须通过以下自检，否则先修再继续：① L2_D01 反映当前架构 ② init 脚本引用的 templates/ 文件全部存在 ③ project-graph relations 不为空（有代码时）④ AGENTS/SKILL/instructions 关键规则一致。agentprecept 自身必须吃自己的狗粮
- 记忆自动生长：Agent 在对话中发现新的用户偏好、项目约束、踩坑教训时，同一 turn 内立即追加到
docs/MEMORY.md对应小节。追加前先 grep 确认无相似条目（不凭记忆判断）；已有条目可更新（替换旧行），不可重复追加。追加后读回验证格式正确。用户说"记住"→立即追加。会话结束前回顾确认全部已写入。模板注释（<!-- ... -->）仅第一次追加时删除 - MCP 缺失检测：当项目已有 AGENTS.md 但
mcp_agentprecept_*tools 不可用时，Agent 应提示用户配置 MCP（agentprecept setup或手动编辑 mcp.json） - Hook/CI 检测：每次会话启动时检查
.git/hooks/pre-commit和.github/workflows/是否存在。缺失 → 提示agentprecept hooks install或agentprecept init --ci。CI gate 是唯一不可跳过的硬拦截——没装等于没有防线 - 模板外脑：docs/ 只有核心 8 文件。需要 API 契约、测试策略、部署方案等模板时，Agent 应知道从 agentprecept 仓库的
templates/（37 个）和methodology/（16 篇）按需取用 - 反馈提示：HANDOFF 状态为 [CLOSING] 时，Agent 应提示："如果这次体验有用，花 2 分钟在 GitHub 填反馈模板"
- 分支纪律：涉及架构/重命名/多文件（>10 文件）的变更必须在独立分支上执行。pre-commit hook 强制拦截 main/master 分支上 >10 文件的 commit（可
--no-verify跳过）；CI gate 维度 11 不可跳过 - 自选维度检查：每 5 轮对话检查自选维度（用户旅程/定位审计/复用/社区就绪度）是否退化。
agentprecept audit --gate报告末尾有自选清单 - 版本规则：严格 semver。PATCH（v0.4.1）修 bug 不增删功能——累积 3+ 修复或 1 个关键修复后发布；MINOR（v0.5.0）新增向后兼容功能，每个功能独立发版不等累积；MAJOR（v1.0.0）破坏性变更。升级方式：全量（
pip install -U agentprecept），11KB wheel 不搞增量 - 发版铁律：任何版本发布到 GitHub/PyPI 前，必须完成 audit --gate FAIL 0 验证 + 将验证结果展示给用户 + 等待用户明确确认（"发"/"发布"/"publish"/"go ahead"）。Agent 严禁未经用户确认私自打 tag、push、或 twine upload。这条规则高于任何自动化——即使 Auto-Pilot 也不得跳过
- 批量创建文件：> 15 个文件时 → 分子代理并行（单个子代理 ≤ 15 文件，超时 180s。依据：世界模拟器 37 文件迁移时 230s 超时）
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 289 lines · 5,095 tokens per session scan A 40cb1dce26e9
AgentPrecept AGENTS.md is an instructions file published in the GitHub repository rg8diaoa/AgentPrecept (9 stars, last pushed 2mo ago), licensed MIT. It adds 5,095 tokens to every session, about $0.0255 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
veritas-kanban AGENTS.md
Instructions for BradGroux/veritas-kanban, covering agents.md — canonical agent instructions for veritas kanban, runtime requirements, repository layout, essential commands and install.
single-file-wbs CLAUDE.md
Instructions for piguo45/single-file-wbs, covering wbs ツール - claude.md, 製品ビジョン・設計指針（#67）, 取り扱いの原則（重要）, アーカイブ運用（過去タスクの整理） and 動作環境.
STAR AGENTS.md
AGENTS.md instructions for wanghao9610/STAR, covering agent instructions, 1. think before coding, 2. simplicity first, 3. surgical changes and 4. goal-driven execution.
AllBeads CLAUDE.md
Instructions for thrashr888/AllBeads, covering claude.md, project overview, what allbeads does, current state (v0.2.0) and rust development commands.
winx-code-agent AGENTS.md
Instructions for gabrielmaialva33/winx-code-agent, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.
jjj AGENTS.md
Instructions for doug/jjj, covering agent instructions, before you write anything, read this first, house rules and verifying changes to this repository.